Estimating microscopy-informed fibre orientations from dMRI data in the UK Biobank

Motivation: There is an unmet need for non-invasive methods to map complex neuroanatomy at the meso-scale. Diffusion MRI (dMRI) tractography is effective but limited by inaccuracies in fibre orientation estimates. Goal(s): To leverage co-aligned microscopy and dMRI from a mostly postmortem dataset to develop a method for high-resolution, hybridised tractography in vivo. Approach: We adapted a network to predict microscopy-informed fibre orientations from dMRI for in vivo data. Trained on a macaque dataset with MRI and microscopy, the network was applied to UK BioBank data. Results: Microscopy-informed fibre orientations aligned with neuroanatomical expectations, producing tractography that captured individual connectivity differences. Impact: We present a deep-learning model that reconstructs microscopy-informed fibre orientations from in vivo human dMRI data, enabling tractography that may better capture individual brain connectivity differences compared to conventional methods.